User Experience Design: A brief analysis of user classification

Speaking of network products, the topic is inseparable from the user, just like the traditional industry consumers. People are complex, user behavior is more complex, users and users are not the same, or that each user is different. A successful internet product often does not meet the needs of all users, but the accurate positioning of a certain type of users and well meet the needs of those users. What kind of users we need to think about, so we need user classification.

Do not classify bad positioning, good user classification Let me know who I am in pursuit of who, meet who, affect who. But not good class and will dislocation, even worse, that how can a product of the user group for reasonable classification, the following to talk about my user classification of some views.

It is generally mentioned that there are several types of users of a particular product that may mainly include the following: High-end and low-end users, student and white-collar users, one or two-line cities and three or four-line city users, active and inactive users, members and non-members, red and non-red drilling users, it and non-it users, primary users, ordinary users, advanced users, etc. , the common feature of these user classifications is to divide users from one dimension or two dimensions, as in Cooper's about Face 2.0, which refers to two user metrics: the Business domain level (domain knowledge) and the computer skill level, Thus dividing the user into the primary user, the ordinary user, the advanced user, simpler to say this method's user classification mode is based on the operation frequency, this classification method may apply in any product, but this kind of user classification actual application effect?

First of all, how to judge a product's user classification effect, mainly from two angles to judge: the Reliability and validity of classification, that is, classification accuracy and accuracy. The accuracy of the classification is that after the class is divided, is not the reality of each user can be positioned to reflect the user's category, that is, any user can be affixed to him a category of labels, and the accuracy of classification refers to the extent to which the resulting user class reflects the actual user contains the attribute meaning, That is to describe the degree to which the characteristic information of each type of user is consistent with all the attributes of the actual user. Accuracy and accuracy are not always perfect at the same time in the actual classification, when you pursue the accuracy of 100% precision will certainly fall, such as only the use of sex to divide users, accuracy is high but not precision, so in the actual user classification to find the accuracy and accuracy of a balance point, to achieve their own classification purposes.

And back to the previous mentioned the division of users into the primary users, ordinary users, advanced users, this division of the method is very high accuracy, but not enough precision, each user can be judged according to the actual situation for the primary users, ordinary users or advanced users, However, the description of the user's characteristics is rarely only two dimensions of operating frequency and computer skill level. This is very inaccurate, in the actual situation, the user's characteristic information contains a lot of users, any one of the characteristics of the different factors will lead to different users to use a product of the behavior habits of preference. For example, the user's age, sex, education, income level, computer level, occupation, geographical, network age and the use of a product target and other factors will lead to different users of different use habits and preferences. So in the user classification, we need to consider how to divide the users from the characteristics of multiple dimensions. What's the actual operation?

First of all, consider the user classification of a product needs which feature factors, that is, from which the number of dimensions to divide users. It is generally considered from the following dimensions: demographic information of the user, the user's computer background (including the user's Internet use background), Internet location, income level, occupation, geography, user experience and preference for the product, what kind of products used, what is the purpose of use, think which is the best use, Factors that affect the selection of a product, through which way to know, the use of product attitude, the use of product specific behavior and other factors. That specific to a product should choose which factors to divide the user, the solution is to first put all the dimensions are listed, and then for these dimensions to conduct user interviews, through the interview can get about the similarities and differences between users. Then all factors were transformed into questionnaires, and the survey data were obtained through scientific sampling questionnaires. The user classification can be obtained by clustering analysis of these user data. What do you need to pay attention to for user clustering?

There are many factors in cluster analysis which affect the final user classification results, and the factors that affect them are: Clustering method selection, distance algorithm selection, clustering variable selection, user class number selection. For clustering methods and distance selection, I tend to recommend the selection of two-step clustering method and log-likelihood distance algorithm, because the user's demographic characteristics and the use of a product behavior preferences are generally classified variables, with the Euclidean distance algorithm, its distance formula expressed by the meaning of the actual meaning is difficult to describe, Or that its distance value in reality is not practical meaning. Clustering variables can choose the interview to get different characteristics of the major factors, but these variables are also related to, and specifically through the constant attempt to adjust, mainly to remove a variable after the clustering results are large differences, if the variable is an important variable, The number of user classes can be determined by the actual situation of the descriptive judgment factors obtained by the clustering and the interview.

How to classify users, how far to subdivide, there is no way to have a pattern or method to generic. So when it comes to the user classification of a specific product, first make clear what you want to classify, and then you need to use these classes after you have finished sorting the class. When it is possible to get a clear product user base and product positioning from the user classification, the classification is basically effective.